Evolution of MLOps in Cian

Currently, there are more than 60 ML models in Cian. Cian uses automated pipelines for making changes to the code of ML models, delivering models to the production, and regular retraining on a schedule. But it wasn't always like that. In the talk, we will learn how Cian came from desperate manual processes of working with models to the ML platform. We will discuss why Cian uses their own ML Model Server for online inference, and how the Jira workflow simplified and accelerated working with models.